On 5/26/25 09:08, Julian Ganz wrote:
May 25, 2025 at 10:56 PM, "Julian Ganz" wrote:
Richard Henderson wrote:
On 5/19/25 16:19, Julian Ganz wrote:
+QEMU_DISABLE_CFI
+static void plugin_vcpu_cb__discon(CPUState *cpu,
+ enum qemu_plugin_discon_type type,
+ uint64_t from)
+{
+ struct qemu_plugin_cb *cb, *next;
+ enum qemu_plugin_event ev;
+ uint64_t to = cpu->cc->get_pc(cpu);
+
+ if (cpu->cpu_index < plugin.num_vcpus) {
+ switch (type) {
+ case QEMU_PLUGIN_DISCON_INTERRUPT:
+ ev = QEMU_PLUGIN_EV_VCPU_INTERRUPT;
+ break;
+ case QEMU_PLUGIN_DISCON_EXCEPTION:
+ ev = QEMU_PLUGIN_EV_VCPU_EXCEPTION;
+ break;
+ case QEMU_PLUGIN_DISCON_HOSTCALL:
+ ev = QEMU_PLUGIN_EV_VCPU_HOSTCALL;
+ break;
No point passing in QEMU_PLUGIN_DISCON_* only to covert it to QEMU_PLUGIN_EV_*.
It easily looks that way, and I myself stubled upon this at least one or two
times, but `type` is the enum we pass to the callback a few lines down and part
of the public plugin API. `ev` on the other hand is the offset in the `cb_list`.
So some translation is neccessary, unfortunately.
I just realized that you probably meant that we should pass ev as a
parameter to plugin_vcpu_cb__discon. This we can obviously do.
Yes, that's what I meant.
r~
